Abstract
More than half of mammalian genes generate multiple messenger RNA isoforms that differ in their 3' untranslated regions (3' UTRs) and therefore in regulatory sequences(1), often associated with cell proliferation and cancer(2,3); however, the mechanisms coordinating alternative 3'-UTR processing for specific mRNA populations remain poorly defined. Here we report that the cytoplasmic-polyadenylation element binding protein 1 (CPEB1), an RNA-binding protein that regulates mRNA translation(4), also controls alternative 3'-UTR processing. CPEB1 shuttles to the nudeus(5,6), where it co-localizes with splicing factors and mediates shortening of hundreds of mRNA 3' UTRs, thereby modulating their translation efficiency in the cytoplasm. CPEB1-mediated 3'-UTR shortening correlates with cell proliferation and tumorigenesis. CPEB1 binding to pre-mRNAs not only directs the use of alternative polyadenylation sites, but also changes alternative splicing by preventing U2AF65 recruitment. Our results reveal a novel function of CPEB1 in mediating alternative 3'-UTR processing, which is coordinated with regulation of mRNA translation, through its dual nuclear and cytoplasmic functions.

Abstract
The adoption of third-party machine learning (ML) cloud services is highly dependent on the security guarantees and the performance penalty they incur on workloads for model training and inference. This paper explores security/performance trade-offs for the distributed Apache Spark framework and its ML library. Concretely, we build upon a key insight: in specific deployment settings, one can reveal carefully chosen non-sensitive operations (e.g. statistical calculations). This allows us to considerably improve the performance of privacy-preserving solutions without exposing the protocol to pervasive ML attacks. In more detail, we propose Soteria, a system for distributed privacy-preserving ML that leverages Trusted Execution Environments (e.g. Intel SGX) to run computations over sensitive information in isolated containers (enclaves). Unlike previous work, where all ML-related computation is performed at trusted enclaves, we introduce a hybrid scheme, combining computation done inside and outside these enclaves. The experimental evaluation validates that our approach reduces the runtime of ML algorithms by up to 41% when compared to previous related work. Our protocol is accompanied by a security proof and a discussion regarding resilience against a wide spectrum of ML attacks.